Skills

LPM provides an AI coding agent skill that helps you build, improve, and monetize packages. The skill understands LPM's package format, quality system, distribution modes, and marketplace.

Install it via skills.sh and use it with your preferred AI coding agent (Claude Code, Codex, VSCode, Cursor, Windsurf, etc.).

LPM Guide

One skill for the entire LPM package lifecycle. It routes to the right workflow based on what you ask, loading only the relevant instructions to keep things fast.

npx skills add lpm-dev/lpm-guide
# or
npx skills add https://github.com/lpm-dev/lpm-guide

View on skills.sh

Workflows

Workflow	What it does	Trigger phrases
Scaffold	Create a new LPM package from scratch with quality-optimized defaults	"create package", "new package", "lpm init"
Migrate	Move an existing npm package to LPM (full or dual publishing)	"migrate to lpm", "npm to lpm", "dual publish"
Improve	Run 28 quality checks + deep analysis, offer automated fixes	"improve quality", "quality score", "prepare for publish"
Source Config	Generate lpm.config.json for configurable source packages	"lpm config", "make configurable", "lpm.config.json"
Distribution	Choose between Private, Pool, and Marketplace modes	"distribution mode", "pool vs marketplace", "monetize"
Pricing	Design Marketplace pricing tiers and license types	"pricing strategy", "how much to charge", "license type"
Deploy	Configure CI/CD for installing or publishing LPM packages	"deploy setup", "CI setup", "build fails 401", "LPM_TOKEN"
Review	Evaluate a package before installing it	"review package", "is this safe", "compare packages"
Skills	Create Agent Skills for AI coding assistants that use your package	"create skills", "agent skills", "lpm skills"

Scaffold

Creates a complete package structure with package.json, README, CHANGELOG, LICENSE, and tests. Supports utility libraries, React components, CLI tools, and source packages. Detects your project context (TypeScript, monorepo, package manager) and scores 70+ out of the box.

What you'll see: Ask "create a new utility package" and the agent will generate your full project structure, wire up exports, create a test file, and configure package.json with correct LPM naming. You'll get something like:

my-package/
├── src/index.ts
├── src/__tests__/index.test.ts
├── package.json          # @lpm.dev/you.my-package, exports, types
├── tsconfig.json
├── README.md             # Install + usage + API docs
├── CHANGELOG.md
└── LICENSE

Migrate

Renames to @lpm.dev/owner.package-name format, updates package.json and README install instructions, configures .npmrc and CI/CD. Supports full migration and dual publishing (npm + LPM) for gradual transitions.

What you'll see: Ask "migrate this package to lpm" and the agent renames your package, updates all references in README and package.json, configures .npmrc for publishing, and optionally sets up dual publishing so the same code goes to both npm and LPM.

Improve

Runs the same 28 checks as lpm publish --check, plus deeper analysis of README completeness vs actual exports, test coverage gaps, dependency audit (misplaced deps, lighter alternatives, license compatibility), and bundle size. Can auto-generate missing READMEs, CHANGELOGs from git history, type definitions, CI config, and tests.

What you'll see: Ask "improve my package quality" and the agent runs all 28 checks, shows your score breakdown by category, then offers to fix issues automatically. Example output:

Quality Score: 62/100 (Fair)

Documentation (14/22):  ✓ README  ✗ Install section  ✗ Usage examples  ✓ CHANGELOG  ✓ LICENSE
Code Quality  (21/31):  ✓ Types  ✓ ESM  ✗ Exports map  ✗ Source maps  ✓ No eval
Testing       (7/11):   ✓ Test files  ✗ Test script  ✗ CI config
Health        (20/36):  ✓ Description  ✗ Keywords  ✓ Repository  ✗ Homepage

I can fix 5 of these automatically. Should I add install/usage sections
to README, create an exports map, add a test script, and set up GitHub Actions?

Source Config

Analyzes your package structure and generates lpm.config.json with config schema, conditional file includes, conditional dependencies, and smart import rewriting.

Distribution

Analyzes your package type, audience, and revenue goals to recommend Private, Pool, or Marketplace. Provides side-by-side comparison with revenue projections. Warns about permanence - Pool and Marketplace modes are irreversible. Checks dependency compatibility before recommending.

What you'll see: Ask "should I use Pool or Marketplace?" and the agent analyzes your package, then presents a comparison:

Based on your package (utility library, 12 exports, MIT license):

            Pool                    Marketplace
Revenue     Passive, per-install    Direct sales, per-seat
Pricing     Included in $12/mo sub  You set the price
Best for    Widely-used utilities   Premium/niche packages
Risk        Lower (volume-based)    Higher (need buyers)
Permanence  ⚠️ Irreversible         ⚠️ Irreversible

Recommendation: Pool — your package is a general utility that benefits
from volume. At 500 monthly installs, estimated Pool earnings: $8-15/mo.

Pricing

Recommends license type (per-seat or per-org), designs tier structure with feature differentiation, and suggests price points based on value-based pricing and competitive analysis. Generates a ready-to-implement pricing table for the LPM dashboard.

Deploy

Configures LPM_TOKEN and .npmrc for Vercel, Netlify, GitHub Actions, GitLab CI, AWS Amplify, Railway, and Docker. Sets up package caching, auto-publish workflows with quality gates, and optionally configures the MCP Server for AI-assisted workflows. Troubleshoots common errors (401, 404, scope-not-found).

What you'll see: Ask "set up CI/CD for my package" and the agent detects your CI platform, generates the workflow file (e.g., .github/workflows/publish.yml), configures .npmrc for LPM auth, and reminds you to set the LPM_TOKEN secret. If you're deploying an app that uses LPM packages, it configures the install step with caching.

Review

Multi-dimensional analysis covering quality score, security (eval, install scripts, network access), maintenance health, API design, and bundle impact. Provides side-by-side package comparison and a clear recommendation: Install, Install with Caution, or Avoid.

Skills

Creates Agent Skills for your package - markdown files in .lpm/skills/ that teach AI coding assistants how to use your package effectively. The agent analyzes your source code, asks clarifying questions about common mistakes and anti-patterns, then generates skill files.

What you'll see: Ask "create skills for my package" and the agent reads your source code and README, asks about common pitfalls, then generates 2-4 skill files in .lpm/skills/:

.lpm/skills/
├── getting-started.md     # Quick start guide for AI assistants
├── best-practices.md      # Recommended patterns
└── anti-patterns.md       # Common mistakes to avoid

Publishing with skills earns up to 10 quality points (has-skills +7, skills-comprehensive +3).

How Skills Work

Skills extend AI coding agents with domain-specific knowledge. After installing a skill, your agent automatically recognizes relevant trigger phrases and loads the skill's instructions and reference materials.

1. Install the skill with npx skills add
2. Open your AI coding agent in your package directory
3. Ask naturally (e.g., "improve my package quality" or "help me set pricing")
4. The skill guides the agent through the analysis and generates fixes

Skills work with any AI coding agent that supports the skills.sh format.